[R-sig-ME] how to get satterthwaite/Kenward-Roger degrees of freedom for heterogeneous-variance one-way ANOVA from an object of nlme::gls

[Lenth, Russell V]

What Rolf says is true, but I'm going to push back a little bit on this. The joint_tests() function in emmeans does not implement SAS's type III F tests, though they are often the same. Joint_tests() results are ALWAYS formulated in terms of estimated marginal means. If the associated EMMs are appropriate, so are the joint tests. 

SAS does the wrong thing with covariates in its type III estimable functions: it sets them to zero instead of to their mean. Consequently, its type III tests are based on comparing intercepts instead of adjusted means. That is an unfortunate mistake that has endured and has generally been misunderstood for about 50 years.

--
Russell V. Lenth  -  Professor Emeritus
Department of Statistics and Actuarial Science  
The University of Iowa  -  Iowa City, IA 52242  USA         
Dept office (319)335-0712  -  FAX (319)335-3017
russell-lenth using uiowa.edu  -  http://www.stat.uiowa.edu/~rlenth/


-----Original Message-----
Date: Fri, 18 Oct 2024 01:53:26 +0000
From: Rolf Turner <rolfturner using posteo.net>
To: "Qiu, Weiliang /US via R-sig-mixed-models"
	<r-sig-mixed-models using r-project.org>
Subject: Re: [R-sig-ME] how to get satterthwaite/Kenward-Roger degrees
	of freedom for heterogeneous-variance one-way ANOVA from an object of
	nlme::gls
Message-ID: <20241018145326.7217405a using elderly-dell>
Content-Type: text/plain; charset="us-ascii"


On Wed, 16 Oct 2024 14:42:24 +0000
"Qiu, Weiliang /US via R-sig-mixed-models"
<r-sig-mixed-models using r-project.org> wrote:

> Greetings. I am using nlme::gls() to perform heterogeneous-variance 
> one-way ANOVA and would like to get type III anova table.

<SNIP>

One should always think very carefully about what hypotheses are actually being tested in a "type III analysis".  Such an analysis may well be "meaningful" but the hypotheses being tested may not be of any real interest. They involve taking averages of cell means over the levels of factors and these averages may not describe anything pertinent.

cheers,

Rolf Turner

--
Honorary Research Fellow
Department of Statistics
University of Auckland
Stats. Dep't. (secretaries) phone:
         +64-9-373-7599 ext. 89622
Home phone: +64-9-480-4619